% Options for packages loaded elsewhere
\PassOptionsToPackage{unicode}{hyperref}
\PassOptionsToPackage{hyphens}{url}
%
\documentclass[
]{article}
\usepackage{lmodern}
\usepackage{amssymb,amsmath}
\usepackage{ifxetex,ifluatex}
\ifnum 0\ifxetex 1\fi\ifluatex 1\fi=0 % if pdftex
  \usepackage[T1]{fontenc}
  \usepackage[utf8]{inputenc}
  \usepackage{textcomp} % provide euro and other symbols
\else % if luatex or xetex
  \usepackage{unicode-math}
  \defaultfontfeatures{Scale=MatchLowercase}
  \defaultfontfeatures[\rmfamily]{Ligatures=TeX,Scale=1}
\fi
% Use upquote if available, for straight quotes in verbatim environments
\IfFileExists{upquote.sty}{\usepackage{upquote}}{}
\IfFileExists{microtype.sty}{% use microtype if available
  \usepackage[]{microtype}
  \UseMicrotypeSet[protrusion]{basicmath} % disable protrusion for tt fonts
}{}
\makeatletter
\@ifundefined{KOMAClassName}{% if non-KOMA class
  \IfFileExists{parskip.sty}{%
    \usepackage{parskip}
  }{% else
    \setlength{\parindent}{0pt}
    \setlength{\parskip}{6pt plus 2pt minus 1pt}}
}{% if KOMA class
  \KOMAoptions{parskip=half}}
\makeatother
\usepackage{xcolor}
\IfFileExists{xurl.sty}{\usepackage{xurl}}{} % add URL line breaks if available
\IfFileExists{bookmark.sty}{\usepackage{bookmark}}{\usepackage{hyperref}}
\hypersetup{
  pdftitle={15B-create\_main\_figure\_captions},
  hidelinks,
  pdfcreator={LaTeX via pandoc}}
\urlstyle{same} % disable monospaced font for URLs
\usepackage[margin=1in]{geometry}
\usepackage{graphicx,grffile}
\makeatletter
\def\maxwidth{\ifdim\Gin@nat@width>\linewidth\linewidth\else\Gin@nat@width\fi}
\def\maxheight{\ifdim\Gin@nat@height>\textheight\textheight\else\Gin@nat@height\fi}
\makeatother
% Scale images if necessary, so that they will not overflow the page
% margins by default, and it is still possible to overwrite the defaults
% using explicit options in \includegraphics[width, height, ...]{}
\setkeys{Gin}{width=\maxwidth,height=\maxheight,keepaspectratio}
% Set default figure placement to htbp
\makeatletter
\def\fps@figure{htbp}
\makeatother
\setlength{\emergencystretch}{3em} % prevent overfull lines
\providecommand{\tightlist}{%
  \setlength{\itemsep}{0pt}\setlength{\parskip}{0pt}}
\setcounter{secnumdepth}{-\maxdimen} % remove section numbering

\title{15B-create\_main\_figure\_captions}
\author{}
\date{\vspace{-2.5em}}

\begin{document}
\maketitle

Replication of BG analyses using both original and new data. The left
panel shows county-level analyses; the right panel shows agency-level
analyses. The BG estimates correspond to their models and can be matched
back to their Table 2 results. The BG results suggest a statistically
significant negative impact of lagged total military aid on five out of
seven crime outcomes (total crime rate, robbery, assault, larceny, and
vehicle theft). However, our county-level replication results suggest a
statistically significant negative impact of lagged total military aid
on only two crime outcomes (robbery and burglary). Most importantly, our
agency-level replication results suggest no statistically significant
impact of lagged total military aid on any crime outcomes. For BG, total
crime rate: \(\beta\) = -59.29, P \textless{} 0.001, and 95\% CI =
{[}-87.1, -31.48{]}; homicide: \(\beta\) = -0.06, P = 0.506, and 95\% CI
= {[}-0.25, 0.12{]}; robbery: \(\beta\) = -6.1, P \textless{} 0.001, and
95\% CI = {[}-8.26, -3.95{]}; assault: \(\beta\) = -5.31, P = 0.03, and
95\% CI = {[}-9.96, -0.65{]}; burglary: \(\beta\) = -8.75, P = 0.109,
and 95\% CI = {[}-19.26, 1.76{]}; larceny: \(\beta\) = -27.43, P
\textless{} 0.001, and 95\% CI = {[}-41.73, -13.13{]}; and vehicle
theft: \(\beta\) = -11.64, P \textless{} 0.001, and 95\% CI = {[}-17.2,
-6.08{]}. For our county-level replication, total crime rate: \(\beta\)
= -20.81, P = 0.092, and 95\% CI = {[}-44.57, 2.94{]}; homicide:
\(\beta\) = 0.02, P = 0.617, and 95\% CI = {[}-0.05, 0.09{]}; robbery:
\(\beta\) = -0.65, P = 0.005, and 95\% CI = {[}-1.08, -0.22{]}; assault:
\(\beta\) = -1.78, P = 0.241, and 95\% CI = {[}-4.72, 1.16{]}; burglary:
\(\beta\) = -6.9, P = 0.019, and 95\% CI = {[}-12.46, -1.34{]}; larceny:
\(\beta\) = -10.06, P = 0.174, and 95\% CI = {[}-24.37, 4.25{]}; and
vehicle theft: \(\beta\) = -1.16, P = 0.124, and 95\% CI = {[}-2.61,
0.29{]}. For our agency-level replication, total crime rate: \(\beta\) =
7.91, P = 0.639, and 95\% CI = {[}-24.92, 40.73{]}; homicide: \(\beta\)
= -0.07, P = 0.594, and 95\% CI = {[}-0.33, 0.19{]}; robbery: \(\beta\)
= 0.44, P = 0.516, and 95\% CI = {[}-0.88, 1.77{]}; assault: \(\beta\) =
-1.63, P = 0.468, and 95\% CI = {[}-5.99, 2.73{]}; burglary: \(\beta\) =
-7.6, P = 0.066, and 95\% CI = {[}-15.52, 0.33{]}; larceny: \(\beta\) =
8.06, P = 0.457, and 95\% CI = {[}-13.02, 29.14{]}; and vehicle theft:
\(\beta\) = -0.93, P = 0.387, and 95\% CI = {[}-3.02, 1.16{]}. The
Replication regressions are run on County data with 15,683 observations
from 2010-2014 and Agency data with 45,331 observations from 2010-2015.
The BG County data is from 2006-2012 and has 17,822 observations. All
regression specifications control for percent in poverty, logged median
household income, unemployment, logged population, share male, share
Black, share aged 15-19, share aged 20-24, share aged 25-34, and
agency/county and year fixed effects. We removed outliers from the
agency-level database that had total crime rates of over one
million.\textbackslash{}

Replication of HPBM analyses using both original and new data. The left
panel shows county-level analyses; the right panel shows agency-level
analyses. The HPBM estimates correspond to their models and can be
matched back to their Table 8 results. The HPBM results suggest
statistically significant negative impacts of lagged military aid items
and value on three of four crime outcomes (robbery, assault, and vehicle
theft). However, our county-level replication results suggest
statistically significant negative impacts of lagged military aid items
and value only on robbery, and only of aid value on vehicle theft.
Moreover, our agency-level replication results generally suggest no
statistically significant impacts of lagged military aid items and value
on any crime outcomes, except for the impact of aid value on robbery.
For HPBM and aid items, homicide: \(\beta\) = -0.22, P = 0.466, and 95\%
CI = {[}-0.81, 0.37{]}; robbery: \(\beta\) = -15.39, P \textless{}
0.001, and 95\% CI = {[}-21.73, -9.06{]}; assault: \(\beta\) = -145.8, P
\textless{} 0.001, and 95\% CI = {[}-211.6, -79.99{]}; and vehicle
theft: \(\beta\) = -114.51, P \textless{} 0.001, and 95\% CI =
{[}-143.76, -85.25{]}. For HPBM and aid value, homicide: \(\beta\) =
0.35, P = 0.216, and 95\% CI = {[}-0.2, 0.9{]}; robbery: \(\beta\) =
-6.23, P = 0.03, and 95\% CI = {[}-11.85, -0.62{]}; assault: \(\beta\) =
-110.35, P = 0.009, and 95\% CI = {[}-192.97, -27.73{]}; and vehicle
theft: \(\beta\) = -55.94, P = 0.001, and 95\% CI = {[}-90.37, -21.5{]}.
For our county-level replication and aid items, homicide: \(\beta\) =
-0.32, P = 0.54, and 95\% CI = {[}-1.35, 0.71{]}; robbery: \(\beta\) =
-5.63, P = 0.04, and 95\% CI = {[}-11, -0.25{]}; assault: \(\beta\) =
-4.44, P = 0.683, and 95\% CI = {[}-25.74, 16.86{]}; and vehicle theft:
\(\beta\) = 2.19, P = 0.789, and 95\% CI = {[}-13.87, 18.26{]}. For our
county-level replication and aid value, homicide: \(\beta\) = 0.11, P =
0.529, and 95\% CI = {[}-0.22, 0.44{]}; robbery: \(\beta\) = -4.37, P
\textless{} 0.001, and 95\% CI = {[}-6.38, -2.36{]}; assault: \(\beta\)
= -2.63, P = 0.419, and 95\% CI = {[}-9, 3.74{]}; and vehicle theft:
\(\beta\) = -8.07, P = 0.003, and 95\% CI = {[}-13.42, -2.71{]}. For our
agency-level replication and aid items, homicide: \(\beta\) = -1, P =
0.397, and 95\% CI = {[}-3.33, 1.32{]}; robbery: \(\beta\) = 0.7, P =
0.912, and 95\% CI = {[}-11.79, 13.2{]}; assault: \(\beta\) = -4.4, P =
0.812, and 95\% CI = {[}-40.63, 31.82{]}; and vehicle theft: \(\beta\) =
0.2, P = 0.987, and 95\% CI = {[}-24.67, 25.07{]}. For our agency-level
replication and aid value, homicide: \(\beta\) = -0.01, P = 0.982, and
95\% CI = {[}-0.62, 0.61{]}; robbery: \(\beta\) = -3.81, P = 0.016, and
95\% CI = {[}-6.89, -0.72{]}; assault: \(\beta\) = 0.53, P = 0.898, and
95\% CI = {[}-7.55, 8.61{]}; and vehicle theft: \(\beta\) = -3.89, P =
0.3, and 95\% CI = {[}-11.24, 3.46{]}. The Replication regressions are
run on County data with 15,460 observations from 2010-2014 and Agency
data with 45,055 observations from 2010-2015. The HPBM County data is
from 2000-2013 and has 36,671 observations. All regression
specifications control for lagged arrest rates, economic controls, and
agency/county and year fixed effects. We removed outliers from the
agency-level database that had total crime rates of over one million.

\end{document}
